ArticlesIs there a Gulf War syndrome?
Introduction
In our first paper (see pages 169–78), we reported that UK military veterans of the Gulf War have higher frequencies of self-reported impaired physical functioning, psychological morbidity, and poor perception of general health than UK servicemen who were not deployed to the Gulf War. In addition, more Gulf War veterans reported non-specific physical symptoms. We investigated whether these findings constitute a new disorder.
For an illness to be recognised as a new disorder it must be sufficiently different from other recognised disorders.1 There is no formal process to investigate whether a set of symptoms are unique to a new illness. Several different methods can be used.
In our first paper we used the empirical approach with epidemiological evidence. This method may, however, miss rare symptoms that could be specific markers for the illness. Assumptions that common symptoms are clinically important may introduce systematic biases and result in misleading theories about possible biological mechanisms.
An alternative method is an analytical approach by factor analysis. Factor analysis is a generic term for several procedures that aim to identify whether the correlations between a set of observed variables can be explained by a few latent, unobserved variables (factors). Exploratory factor analysis is used for preliminary investigations of a set of observed variables. In a population with diverse symptoms, this method makes no a-priori assumption about the composition of the factors, can potentially identify clinically important factors, and commonly provides a concise description of the data. Specifically, relevant factors, and the variables that load on to them, can be used to suggest a structure that can be tested on a new set of data in confirmatory factor analysis.2 For confirmatory factor analysis, a specific factor structure is assumed which, after estimation of its parameters, leads to predicted values for the correlations between the observed variables. Whether the specified structure provides an adequate explanation of how the observed variables fit, shown by correlations between them, is determined by how close the predicted correlations are to those observed. Judgment of the fit of models in confirmatory factor analysis is rarely straightforward and various measures of fit are generally used.
Factor analysis has been used in two studies of Gulf-War-related illnesses. Haley and colleagues3 did an exploratory factor analysis of 52 symptoms in 249 Gulf War veterans selected from one US reserve unit. Six factors accounted for 71% of the variance of observed variables. These factors were: impaired cognition, confusion-ataxia, arthromyoneuropathy, phobia-apraxia, fever-adenopathy, and weakness-incontinence. They interpreted their findings as evidence for a unique Gulf War syndrome, despite the lack of a comparison population as a control group. Haley and colleagues' findings and interpretations have heightened the controversy as to whether a unique Gulf War syndrome exists4, 5, 6, 7, 8 and have yet to be replicated.
Fukuda and colleagues9 did an exploratory factor analysis of 35 symptoms in a random sample of current service personnel from four US Air Force units. Three factors-mood and cognition, musculoskeletal, and respiratory-accounted for 39% of the common variance. A second factor analysis of the symptoms that loaded on to these factors in a separate random sample from the same population identified two factors that they labelled mood-cognition-fatigue and musculoskeletal. The investigators developed a clinical case definition of Gulf-War-related illnesses based on whether symptoms in the whole sample had a chronic course, were commonly reported, and were more frequent than in the non-Gulf-War sample. This definition identified fatigue, difficulty remembering or concentrating, difficulty sleeping, moodiness, joint pains, and joint stiffness as the criteria for case definition. Agreement was good between factor-derived and clinically derived symptoms which suggests acceptable construct validity of the analytical approach. Ex-servicemen were not, however, included, which might have introduced a healthy worker bias.
We aimed to identify underlying factors that explained the correlations among symptoms reported in UK servicemen deployed in the Gulf War, and to assess, by confirmatory factor analysis, the fit of the factor structure in the Gulf War cohort in servicemen deployed to the Bosnia conflict and in active servicemen who were not deployed to the Gulf War (Era) at the time of the Gulf War. We tested whether the factor structure proposed by Haley and colleagues2 could be replicated in our Gulf War, Bosnia, and Era cohorts.
Section snippets
Methods
We did a population-based cross-sectional postal survey to compare the health profiles of three UK military samples.
Results
The response rates and sociodemographic and health characteristics for all men are reported in our first paper, as well as the 15 most commonly reported physical symptoms. Nearly all symptoms were reported most in the Gulf War cohort and decreased progressively from the Bosnia to the Era cohort.
In exploratory principal factor analysis, ten factors with eigenvalues of more than 1·0 accounted for 42·1% of the variance in the Gulf War cohort. 69 (2·1%) questionnaires had missing values. The
Discussion
The latent dimensions (factors) that underlie the pattern of symptom reporting in the Gulf War cohort seemed to differ little from those in the Bosnia and Era cohorts. Therefore, although the frequency of symptom reporting was higher in the Gulf War cohort, the underlying structure of the correlations between symptoms was similar to that in the other cohorts. This finding seems to provide evidence against the existence of a unique Gulf War syndrome.
Whether the three factors we analysed
References (12)
- et al.
Invited commentary: how would we know a Gulf War Syndrome if we saw one?
Am J Epidemiol
(1997) - et al.
Covariances and latent variables using EQS
(1993) - et al.
Is there a Gulf War Syndrome? Searching for syndromes by factor analysis of symptoms
JAMA
(1997) Point: bias from the 'Healthy-warrior effect' and unequal follow-up in three government studies of health effects of the Gulf War
Am J Epidemiol
(1998)- et al.
Counterpoint: responding to suppositions and misunderstandings
Am J Epidemiol
(1998) - et al.
Counterpoint: responding to inadequate critique of birth defects paper
Am J Epidemiol
(1998)